{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 68219,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.01,
      "learning_rate": 8e-05,
      "loss": 2.273,
      "step": 500
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.999957460493864e-05,
      "loss": 2.2497,
      "step": 1000
    },
    {
      "epoch": 0.02,
      "learning_rate": 7.999829842880257e-05,
      "loss": 2.3728,
      "step": 1500
    },
    {
      "epoch": 0.03,
      "learning_rate": 7.999617149873574e-05,
      "loss": 2.3576,
      "step": 2000
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.999319385997746e-05,
      "loss": 2.3374,
      "step": 2500
    },
    {
      "epoch": 0.04,
      "learning_rate": 7.998936557586135e-05,
      "loss": 2.3538,
      "step": 3000
    },
    {
      "epoch": 0.05,
      "learning_rate": 7.998468672781407e-05,
      "loss": 2.366,
      "step": 3500
    },
    {
      "epoch": 0.06,
      "learning_rate": 7.997915741535355e-05,
      "loss": 2.3321,
      "step": 4000
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.997277775608694e-05,
      "loss": 2.2838,
      "step": 4500
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.996554788570796e-05,
      "loss": 2.2679,
      "step": 5000
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.995746795799422e-05,
      "loss": 2.4041,
      "step": 5500
    },
    {
      "epoch": 0.09,
      "learning_rate": 7.994853814480376e-05,
      "loss": 2.3415,
      "step": 6000
    },
    {
      "epoch": 0.1,
      "learning_rate": 7.99387586360715e-05,
      "loss": 2.3589,
      "step": 6500
    },
    {
      "epoch": 0.1,
      "learning_rate": 7.992812963980518e-05,
      "loss": 2.3203,
      "step": 7000
    },
    {
      "epoch": 0.11,
      "learning_rate": 7.991665138208094e-05,
      "loss": 2.3517,
      "step": 7500
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.990432410703848e-05,
      "loss": 2.4286,
      "step": 8000
    },
    {
      "epoch": 0.12,
      "learning_rate": 7.989114807687589e-05,
      "loss": 2.3227,
      "step": 8500
    },
    {
      "epoch": 0.13,
      "learning_rate": 7.987712357184408e-05,
      "loss": 2.3945,
      "step": 9000
    },
    {
      "epoch": 0.14,
      "learning_rate": 7.98622508902408e-05,
      "loss": 2.3775,
      "step": 9500
    },
    {
      "epoch": 0.15,
      "learning_rate": 7.984653034840432e-05,
      "loss": 2.3522,
      "step": 10000
    },
    {
      "epoch": 0.15,
      "learning_rate": 7.982996228070671e-05,
      "loss": 2.334,
      "step": 10500
    },
    {
      "epoch": 0.16,
      "learning_rate": 7.981254703954664e-05,
      "loss": 2.3742,
      "step": 11000
    },
    {
      "epoch": 0.17,
      "learning_rate": 7.979428499534201e-05,
      "loss": 2.2517,
      "step": 11500
    },
    {
      "epoch": 0.18,
      "learning_rate": 7.977517653652199e-05,
      "loss": 2.3994,
      "step": 12000
    },
    {
      "epoch": 0.18,
      "learning_rate": 7.975522206951876e-05,
      "loss": 2.3606,
      "step": 12500
    },
    {
      "epoch": 0.19,
      "learning_rate": 7.973442201875895e-05,
      "loss": 2.3634,
      "step": 13000
    },
    {
      "epoch": 0.2,
      "learning_rate": 7.971277682665446e-05,
      "loss": 2.3061,
      "step": 13500
    },
    {
      "epoch": 0.21,
      "learning_rate": 7.969028695359319e-05,
      "loss": 2.4985,
      "step": 14000
    },
    {
      "epoch": 0.21,
      "learning_rate": 7.966695287792921e-05,
      "loss": 2.3908,
      "step": 14500
    },
    {
      "epoch": 0.22,
      "learning_rate": 7.96427750959725e-05,
      "loss": 2.3605,
      "step": 15000
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.961775412197857e-05,
      "loss": 2.347,
      "step": 15500
    },
    {
      "epoch": 0.23,
      "learning_rate": 7.959189048813735e-05,
      "loss": 2.3233,
      "step": 16000
    },
    {
      "epoch": 0.24,
      "learning_rate": 7.95651847445619e-05,
      "loss": 2.3415,
      "step": 16500
    },
    {
      "epoch": 0.25,
      "learning_rate": 7.953763745927682e-05,
      "loss": 2.4679,
      "step": 17000
    },
    {
      "epoch": 0.26,
      "learning_rate": 7.950924921820606e-05,
      "loss": 2.443,
      "step": 17500
    },
    {
      "epoch": 0.26,
      "learning_rate": 7.948002062516052e-05,
      "loss": 2.5141,
      "step": 18000
    },
    {
      "epoch": 0.27,
      "learning_rate": 7.944995230182513e-05,
      "loss": 2.4339,
      "step": 18500
    },
    {
      "epoch": 0.28,
      "learning_rate": 7.941904488774571e-05,
      "loss": 2.4308,
      "step": 19000
    },
    {
      "epoch": 0.29,
      "learning_rate": 7.938729904031533e-05,
      "loss": 2.4312,
      "step": 19500
    },
    {
      "epoch": 0.29,
      "learning_rate": 7.93547154347603e-05,
      "loss": 2.4756,
      "step": 20000
    },
    {
      "epoch": 0.3,
      "learning_rate": 7.932129476412592e-05,
      "loss": 2.4426,
      "step": 20500
    },
    {
      "epoch": 0.31,
      "learning_rate": 7.928703773926155e-05,
      "loss": 2.4779,
      "step": 21000
    },
    {
      "epoch": 0.32,
      "learning_rate": 7.925194508880567e-05,
      "loss": 2.4671,
      "step": 21500
    },
    {
      "epoch": 0.32,
      "learning_rate": 7.921601755917029e-05,
      "loss": 2.4473,
      "step": 22000
    },
    {
      "epoch": 0.33,
      "learning_rate": 7.917925591452508e-05,
      "loss": 2.3929,
      "step": 22500
    },
    {
      "epoch": 0.34,
      "learning_rate": 7.914166093678117e-05,
      "loss": 2.4158,
      "step": 23000
    },
    {
      "epoch": 0.34,
      "learning_rate": 7.910323342557442e-05,
      "loss": 2.4607,
      "step": 23500
    },
    {
      "epoch": 0.35,
      "learning_rate": 7.906397419824855e-05,
      "loss": 2.4866,
      "step": 24000
    },
    {
      "epoch": 0.36,
      "learning_rate": 7.902388408983759e-05,
      "loss": 2.3708,
      "step": 24500
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.898296395304824e-05,
      "loss": 2.4718,
      "step": 25000
    },
    {
      "epoch": 0.37,
      "learning_rate": 7.894121465824175e-05,
      "loss": 2.4436,
      "step": 25500
    },
    {
      "epoch": 0.38,
      "learning_rate": 7.889863709341528e-05,
      "loss": 2.498,
      "step": 26000
    },
    {
      "epoch": 0.39,
      "learning_rate": 7.885523216418312e-05,
      "loss": 2.4418,
      "step": 26500
    },
    {
      "epoch": 0.4,
      "learning_rate": 7.881100079375742e-05,
      "loss": 2.3653,
      "step": 27000
    },
    {
      "epoch": 0.4,
      "learning_rate": 7.876594392292848e-05,
      "loss": 2.5256,
      "step": 27500
    },
    {
      "epoch": 0.41,
      "learning_rate": 7.872006251004482e-05,
      "loss": 2.549,
      "step": 28000
    },
    {
      "epoch": 0.42,
      "learning_rate": 7.867335753099278e-05,
      "loss": 2.4543,
      "step": 28500
    },
    {
      "epoch": 0.43,
      "learning_rate": 7.86258299791757e-05,
      "loss": 2.4647,
      "step": 29000
    },
    {
      "epoch": 0.43,
      "learning_rate": 7.857748086549292e-05,
      "loss": 2.5375,
      "step": 29500
    },
    {
      "epoch": 0.44,
      "learning_rate": 7.852831121831812e-05,
      "loss": 2.4895,
      "step": 30000
    },
    {
      "epoch": 0.45,
      "learning_rate": 7.847832208347754e-05,
      "loss": 2.4156,
      "step": 30500
    },
    {
      "epoch": 0.45,
      "learning_rate": 7.842751452422775e-05,
      "loss": 2.5006,
      "step": 31000
    },
    {
      "epoch": 0.46,
      "learning_rate": 7.8375889621233e-05,
      "loss": 2.459,
      "step": 31500
    },
    {
      "epoch": 0.47,
      "learning_rate": 7.83234484725422e-05,
      "loss": 2.469,
      "step": 32000
    },
    {
      "epoch": 0.48,
      "learning_rate": 7.827019219356568e-05,
      "loss": 2.4331,
      "step": 32500
    },
    {
      "epoch": 0.48,
      "learning_rate": 7.821612191705128e-05,
      "loss": 2.487,
      "step": 33000
    },
    {
      "epoch": 0.49,
      "learning_rate": 7.816123879306048e-05,
      "loss": 2.5139,
      "step": 33500
    },
    {
      "epoch": 0.5,
      "learning_rate": 7.810554398894376e-05,
      "loss": 2.5117,
      "step": 34000
    },
    {
      "epoch": 0.51,
      "learning_rate": 7.804903868931584e-05,
      "loss": 2.5537,
      "step": 34500
    },
    {
      "epoch": 0.51,
      "learning_rate": 7.79917240960305e-05,
      "loss": 2.4394,
      "step": 35000
    },
    {
      "epoch": 0.52,
      "learning_rate": 7.7933601428155e-05,
      "loss": 2.4285,
      "step": 35500
    },
    {
      "epoch": 0.53,
      "learning_rate": 7.78746719219441e-05,
      "loss": 2.5693,
      "step": 36000
    },
    {
      "epoch": 0.54,
      "learning_rate": 7.781493683081388e-05,
      "loss": 2.4932,
      "step": 36500
    },
    {
      "epoch": 0.54,
      "learning_rate": 7.775439742531495e-05,
      "loss": 2.4468,
      "step": 37000
    },
    {
      "epoch": 0.55,
      "learning_rate": 7.769305499310553e-05,
      "loss": 2.4614,
      "step": 37500
    },
    {
      "epoch": 0.56,
      "learning_rate": 7.763091083892402e-05,
      "loss": 2.6549,
      "step": 38000
    },
    {
      "epoch": 0.56,
      "learning_rate": 7.756796628456121e-05,
      "loss": 2.4871,
      "step": 38500
    },
    {
      "epoch": 0.57,
      "learning_rate": 7.750422266883222e-05,
      "loss": 2.413,
      "step": 39000
    },
    {
      "epoch": 0.58,
      "learning_rate": 7.743968134754806e-05,
      "loss": 2.3566,
      "step": 39500
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.737434369348664e-05,
      "loss": 2.4936,
      "step": 40000
    },
    {
      "epoch": 0.59,
      "learning_rate": 7.730821109636379e-05,
      "loss": 2.453,
      "step": 40500
    },
    {
      "epoch": 0.6,
      "learning_rate": 7.724128496280346e-05,
      "loss": 2.4851,
      "step": 41000
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.717356671630802e-05,
      "loss": 2.4564,
      "step": 41500
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.710505779722786e-05,
      "loss": 2.5133,
      "step": 42000
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.703575966273073e-05,
      "loss": 2.4546,
      "step": 42500
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.696567378677089e-05,
      "loss": 2.5157,
      "step": 43000
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.689480166005756e-05,
      "loss": 2.4248,
      "step": 43500
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.682314479002344e-05,
      "loss": 2.4853,
      "step": 44000
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.67507047007924e-05,
      "loss": 2.4615,
      "step": 44500
    },
    {
      "epoch": 0.66,
      "learning_rate": 7.667748293314729e-05,
      "loss": 2.5391,
      "step": 45000
    },
    {
      "epoch": 0.67,
      "learning_rate": 7.6603481044497e-05,
      "loss": 2.4464,
      "step": 45500
    },
    {
      "epoch": 0.67,
      "learning_rate": 7.652870060884345e-05,
      "loss": 2.4941,
      "step": 46000
    },
    {
      "epoch": 0.68,
      "learning_rate": 7.645314321674803e-05,
      "loss": 2.4708,
      "step": 46500
    },
    {
      "epoch": 0.69,
      "learning_rate": 7.637681047529781e-05,
      "loss": 2.5972,
      "step": 47000
    },
    {
      "epoch": 0.7,
      "learning_rate": 7.629970400807136e-05,
      "loss": 2.5369,
      "step": 47500
    },
    {
      "epoch": 0.7,
      "learning_rate": 7.622182545510419e-05,
      "loss": 2.4348,
      "step": 48000
    },
    {
      "epoch": 0.71,
      "learning_rate": 7.61431764728539e-05,
      "loss": 2.4546,
      "step": 48500
    },
    {
      "epoch": 0.72,
      "learning_rate": 7.606375873416491e-05,
      "loss": 2.4378,
      "step": 49000
    },
    {
      "epoch": 0.73,
      "learning_rate": 7.598357392823292e-05,
      "loss": 2.471,
      "step": 49500
    },
    {
      "epoch": 0.73,
      "learning_rate": 7.590262376056896e-05,
      "loss": 2.4677,
      "step": 50000
    },
    {
      "epoch": 0.74,
      "learning_rate": 7.58209099529631e-05,
      "loss": 2.4245,
      "step": 50500
    },
    {
      "epoch": 0.75,
      "learning_rate": 7.573843424344783e-05,
      "loss": 2.4734,
      "step": 51000
    },
    {
      "epoch": 0.75,
      "learning_rate": 7.565519838626113e-05,
      "loss": 2.4158,
      "step": 51500
    },
    {
      "epoch": 0.76,
      "learning_rate": 7.557120415180916e-05,
      "loss": 2.5098,
      "step": 52000
    },
    {
      "epoch": 0.77,
      "learning_rate": 7.548645332662853e-05,
      "loss": 2.5478,
      "step": 52500
    },
    {
      "epoch": 0.78,
      "learning_rate": 7.540094771334835e-05,
      "loss": 2.4502,
      "step": 53000
    },
    {
      "epoch": 0.78,
      "learning_rate": 7.531468913065192e-05,
      "loss": 2.4264,
      "step": 53500
    },
    {
      "epoch": 0.79,
      "learning_rate": 7.522767941323798e-05,
      "loss": 2.4903,
      "step": 54000
    },
    {
      "epoch": 0.8,
      "learning_rate": 7.513992041178174e-05,
      "loss": 2.4486,
      "step": 54500
    },
    {
      "epoch": 0.81,
      "learning_rate": 7.505141399289549e-05,
      "loss": 2.5171,
      "step": 55000
    },
    {
      "epoch": 0.81,
      "learning_rate": 7.496216203908891e-05,
      "loss": 2.5396,
      "step": 55500
    },
    {
      "epoch": 0.82,
      "learning_rate": 7.487216644872901e-05,
      "loss": 2.4514,
      "step": 56000
    },
    {
      "epoch": 0.83,
      "learning_rate": 7.478142913599978e-05,
      "loss": 2.4017,
      "step": 56500
    },
    {
      "epoch": 0.84,
      "learning_rate": 7.468995203086146e-05,
      "loss": 2.4591,
      "step": 57000
    },
    {
      "epoch": 0.84,
      "learning_rate": 7.459773707900946e-05,
      "loss": 2.5764,
      "step": 57500
    },
    {
      "epoch": 0.85,
      "learning_rate": 7.450478624183306e-05,
      "loss": 2.5013,
      "step": 58000
    },
    {
      "epoch": 0.86,
      "learning_rate": 7.441110149637363e-05,
      "loss": 2.51,
      "step": 58500
    },
    {
      "epoch": 0.86,
      "learning_rate": 7.431668483528254e-05,
      "loss": 2.3992,
      "step": 59000
    },
    {
      "epoch": 0.87,
      "learning_rate": 7.422153826677887e-05,
      "loss": 2.4671,
      "step": 59500
    },
    {
      "epoch": 0.88,
      "learning_rate": 7.412566381460662e-05,
      "loss": 2.5362,
      "step": 60000
    },
    {
      "epoch": 0.89,
      "learning_rate": 7.402906351799175e-05,
      "loss": 2.4981,
      "step": 60500
    },
    {
      "epoch": 0.89,
      "learning_rate": 7.39317394315987e-05,
      "loss": 2.4574,
      "step": 61000
    },
    {
      "epoch": 0.9,
      "learning_rate": 7.383369362548674e-05,
      "loss": 2.4777,
      "step": 61500
    },
    {
      "epoch": 0.91,
      "learning_rate": 7.373492818506597e-05,
      "loss": 2.4628,
      "step": 62000
    },
    {
      "epoch": 0.92,
      "learning_rate": 7.363544521105292e-05,
      "loss": 2.4506,
      "step": 62500
    },
    {
      "epoch": 0.92,
      "learning_rate": 7.353524681942585e-05,
      "loss": 2.5644,
      "step": 63000
    },
    {
      "epoch": 0.93,
      "learning_rate": 7.343433514137987e-05,
      "loss": 2.5131,
      "step": 63500
    },
    {
      "epoch": 0.94,
      "learning_rate": 7.333271232328141e-05,
      "loss": 2.5575,
      "step": 64000
    },
    {
      "epoch": 0.95,
      "learning_rate": 7.32303805266227e-05,
      "loss": 2.5901,
      "step": 64500
    },
    {
      "epoch": 0.95,
      "learning_rate": 7.312734192797583e-05,
      "loss": 2.5701,
      "step": 65000
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.302359871894635e-05,
      "loss": 2.5201,
      "step": 65500
    },
    {
      "epoch": 0.97,
      "learning_rate": 7.291915310612666e-05,
      "loss": 2.5402,
      "step": 66000
    },
    {
      "epoch": 0.97,
      "learning_rate": 7.281400731104918e-05,
      "loss": 2.4797,
      "step": 66500
    },
    {
      "epoch": 0.98,
      "learning_rate": 7.2708163570139e-05,
      "loss": 2.4836,
      "step": 67000
    },
    {
      "epoch": 0.99,
      "learning_rate": 7.260162413466636e-05,
      "loss": 2.4707,
      "step": 67500
    },
    {
      "epoch": 1.0,
      "learning_rate": 7.249439127069873e-05,
      "loss": 2.4334,
      "step": 68000
    },
    {
      "epoch": 1.0,
      "eval_bleu": 1.0,
      "eval_brevity_penalty": 1.0,
      "eval_length_ratio": 1.0,
      "eval_loss": 2.243044853210449,
      "eval_precisions": [
        1.0,
        1.0,
        1.0,
        1.0
      ],
      "eval_reference_length": 7761920,
      "eval_runtime": 17988.3779,
      "eval_samples_per_second": 0.843,
      "eval_steps_per_second": 0.421,
      "eval_translation_length": 7761920,
      "step": 68219
    }
  ],
  "logging_steps": 500,
  "max_steps": 341095,
  "num_train_epochs": 5,
  "save_steps": 5000,
  "total_flos": 3.144579296777994e+17,
  "trial_name": null,
  "trial_params": null
}